Table 8 Unified performance comparison of transformer models for routing and urgency estimation.

From: A zero-shot LLM framework for multimodal grievance classification, urgency scoring, and abuse detection in civic feedback systems

Model

Accuracy (%)

MAE

Latency (ms)

GPU Req.

BERT-base (fine-tuned)

94.1

0.052

128

Yes

RoBERTa-base (fine-tuned)

94.8

0.048

141

Yes

DeBERTa-v3-base (fine-tuned)

95.2

0.061

158

Yes

LLaMA-3 8B (zero-shot)

93.7

0.059

420

Yes

MobileBERT (zero-shot, Proposed)

92.4

0.041

19

No

Back to article page